RELACIÓN BENEFICIO / COSTO

The policy evaluation regressions show that there is a significant effect on some subjects’ results of the policy introduction but that it is not universal – the effect is different in the two groups

of LEAs and only significant for the poorer LEAs. Common variation due to the year effects accounts for much of the variation in results, with the remaining local variation insufficient to identify a common policy effect in the simplest regressions. In several cases the policy has a different signed effect between the groups and it appears that these effects counter act each other such that when estimated as a single effect it is not significantly different to zero. In other cases it is only a marginally significant effect in the poorer LEAs and not significant in the better off LEAs such that when estimated as a single parameter the poorer LEAs’ effect is not strong enough to identify a significant overall effect.

In terms of evidence of the policy effects working through the actual take-up rates of ‘other’ sector places themselves, it is only in maths scores that we see the significant policy effects mirrored by the significant ‘other’ sector take-up rate impacts. In the better off LEAs there is a positive (though not significant) policy effect for maths at level 2B or higher and this is echoed and significant in the main sample estimation of the effect of the take-up rate of ‘other’ sector places in these LEAs. For level 3 or higher results the policy evaluation regressions suggest a negative policy effect on results in the poorer LEAs, and this effect is identified in the coefficient on the take-up of ‘other’ sector places in these LEAs, in the more complex model regressions. However, these maths results are sensitive to the choice of sample and do not remain in all of the other robustness checks.

Looking at the more indirect effects, it appears that for both reading and writing the main- tained nursery and primary school settings have a small but significant positive effect on attain- ment at the intermediate level for the poorer LEAs when accessed by 3-year olds. These settings are also associated with a larger and significant positive effect on attainment at the higher level for the better off LEAs when accessed by 4-year olds.

However, for 4-year olds it is not possible to interpret this as a causal effect. The introduction of free early education places for 3-year olds was a shock to the market for childcare and early education for 3-year olds but much less so for 4-year olds. The choice of setting for 4-year olds is likely to be much more influenced by the consideration of where the child will be for their primary school education – the nursery and primary school settings for 4-year olds include the reception class, which children can attend for between one and three terms depending on the choice of the parents and the birthday of the child. Table 2.20 shows the percentage of 4-year olds attending a nursery or primary school that are in an infant or reception class as opposed to

a nursery class, for each region of England for 1998-200250. It is clear that a large majority of 4-year olds in nursery and primary school settings are in fact in an infant or reception class in the primary school – and thus in the school that they will attend for their primary education. Therefore the choice to send the child to a maintained school setting rather than a private setting at age 4 is likely to be strongly influenced by the parents’ preferences over primary schools and therefore it is not appropriate to assume that within the LEA, allocation of 4-year olds to school or ‘other’ sector settings is uncorrelated with unobservables. We would suspect there could be a significant upward bias in the estimated effect: parents with strong preferences over their child’s education are likely to try to get their 4-year old into a reception class in a particular, good, primary school rather than use a free place in a private nursery setting. The children of these parents are also more likely to achieve higher results due to the parental preferences over education, leading to an upward bias in the coefficient on 4-year olds attendance of maintained school places.

It is worth noting that this argument does not apply to the case of 3-year olds since it is policy that attendance at a maintained school as a 3-year old does not have any influence on the probability of gaining a place in the reception intake at that school. Thus the link between early education place setting and primary school place does not exist for 3-year olds in the way that it does for 4-year olds and is therefore much less likely to create a selection problem. In theory the early education places for 3-year olds in the school and in the ‘other’, non-school, settings are identical so a parent with strong educational preferences should view them as equal and not select on this into either particular sector.

The effect of maintained school settings’ attendance by 3-year olds in the poorer LEAs on their reading and writing results is in line with what our prior may be. I am comparing the effect of attendance in one of the settings with staying at home and not attending any setting and it appears that, compared with staying at home, a greater proportion of the LEA’s children attending a maintained nursery setting aged 3 is associated with more children attaining the expected level (i.e. at least level 2B) on their KS1 assessment. If we expect that the average home environment for children in these poorer LEAs is less educationally beneficial to a child than attendance part-time in a nursery school setting, then we would expect that increasing take-up by 3-year olds in these LEAs would be associated with positive effects on child outcomes. Given that reading and writing in particular are skills more readily advanced through nursery school

50

attendance (as opposed to maths skills), we might expect that we would find effects on results in these areas.

If we believe that for the better off LEAs, the quality of home inputs is on average better, then it may be no surprise that attendance in more formal settings aged 3 or 4 does not seem to affect the intermediate level of attainment but attendance at age 4 does have a significant effect on attainment at the higher level – and for maths also. As discussed, there is likely to be an upward bias in the coefficient due to selection. However, it is also consistent with the idea that children who develop skills early can be encouraged to develop further and beyond expectations in the more formal nursery and primary school settings whereas this may be more difficult at home.

It is noticeable that the effects at level 3 are greater in magnitude and relative size than the results at level 2B and again this could be due to an upward bias but is also in line with what our prior may be. Bearing in mind that I am unable to distinguish between no effect at all and an effect that is mitigated by the first two years of formal schooling, it is perhaps to be expected that the first years of schooling would do a better job of mitigating any advantage that nursery attendees had over their home-staying class mates at the intermediate level, to the extent that the measured effect on KS1 results whilst statistically significant in the poorer LEAs, is relatively small. For the higher level results, it may be more likely to be the case that the advantage of early exposure to the formal educational setting remains, if this gives the platform on which to build – with the initial schooling having a greater gap to make up if these children in better off LEAs who attend early education places are in a better position to build on their early advantage. As Table 2.20 shows that a large proportion of 4-year olds in nursery and primary school settings are in fact in an infant or reception class in the primary school, the “result” for 4-year olds suggests that results are better at the higher levels for reading, writing and maths in the better off LEAs when more children start school at age 4 rather than remaining at home aged 4.

It has to be borne in mind that the measurement error issues outlined in section 2.6, would suggest that the estimated coefficients are lower bounds for the estimates of the true parameters. However, even taking this into account, in terms of policy implications conclusions are slightly tempered by the magnitudes of the robust results that I find. As discussed above, if the take-up of early education places in maintained nursery and primary schools, by 3-year olds in the poorer LEAs, could be raised by 10%-points, the effect would be to increase the percentage attaining

level 2B or higher in reading by just 0.6%-points and in writing by 0.7%-points. Over the time of the panel, the mean percentage of children in these poorer LEAs attaining level 2B or higher in reading is 65.68% with a within standard deviation of 1.66%-points - so we can see that even a 10%-point change in take-up is making only a relatively small impact. Similarly the mean attainment in these LEAs for writing level 2B is 56.30% with a within standard deviation of 2.19%-points thus even more so for writing, the impact of increasing take-up even by 10%-points is relatively small.

One way in which to assess these findings is to consider what the estimated effect would be, ceteris paribus, if all of the 3-year olds in the poorer LEAs who attend ‘other’ sector places switched to maintained schools places. At the end of the panel, the median take-up of places in the ‘other’ sector by 3-year olds in the poorer LEAs was 30.29% (mean 28.77%). If the take-up in the maintained schools sector increased by 30%-points the effect on reading level 2B would be to raise results by approximately 1.8%-points which is just over one within standard deviation. For writing level 2B the effect of a 30%-point increase in maintained school place take-up would be an increase of approximately 2.1%-points which is just under one within standard deviation. Therefore this admittedly basic calculation (not taking into account capacity constraints or allowing non-linear effects in increasing take-up) suggests that expanding the maintained schools provision of places for 3-year olds in the poorer areas, in the stead of ‘other’ sector places, would lead to a significant increase in reading and writing results.

One other way in which to assess the size of these significant parameters, is to compare the effect of a 10%-point increase in the take-up rate of free early education places in the maintained schools sector, by 3-year olds in the poorer LEAs, with the effect of reducing the average KS1 class size by one child51.

The effect of increasing take-up by 3-year olds in the poorer LEAs by 10%-points, would for reading level 2B or higher, have an effect equivalent to reducing the average KS1 class size by one child. Put in this context, we can see that though the effect of raising this take-up rate seems very small, it is of the same order as reducing class size by one child. For writing level 2B or higher, the effect of increasing take-up rate by 3-year olds in the poorer LEAs by 10%-points is not quite as great in comparison to reducing the average class size. Such an increase in take- up rate would correspond to a reduction of approximately three-quarters of one child, or put another way, it would take an increase in the take-up rate by 3-year olds in the poorer LEAs, of

51

Clearly this exercise makes the simplifying assumption that there is a monotonic effect of reducing KS1 class sizes, which in reality is not likely to be the case.

approximately 14%-points to have an effect on results equivalent to a reduction by one child in the average KS1 class size. Again, in this context, the seemingly small effects of an increase in 3-year olds in poor LEAs taking their free early education place in maintained schools settings, is seen to be of more substantial size than initially thought.

Moreover, these results are in line with the EPPE study which found that nursery schools and classes were the better pre-school environments for 3- and 4-year olds to attend prior to compulsory schooling, which adds to their credence.